INDEX

Explanations

terms that correspond or are related to specific historical events or concepts

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ghazi

-0.80

uv

-0.77

oufl

-0.69

cloth

-0.67

asel

-0.65

asts

-0.64

asted

-0.63

liquid

-0.62

 helicop

-0.62

agers

-0.61

POSITIVE LOGITS

ingly

1.04

SHIP

0.83

ing

0.80

ificantly

0.77

ences

0.76

 Emin

0.76

ette

0.69

ities

0.69

icut

0.68

MPG

0.67

Activations Density 0.026%

terms that correspond or are related to specific historical events or concepts

No Comments

No Known Activations